Forecasting with Prophet-When Does It Make Sense

Every organization generates data. Sales, customer activity, website traffic, daily operations, and system usage all build up a history over time.

The value of this data is not only in understanding what happened yesterday. It is also in preparing for what may happen next and making better decisions today.

This is where forecasting is useful for business leaders.

Prophet, developed by Meta’s Core Data Science team, is a practical way to forecast business numbers that follow repeating patterns, seasonal changes, and changing growth.

However, Prophet is not suitable for every forecasting problem. How well it works depends on the business problem, the past data available, and the decisions the forecast will support.

Where Prophet Can Help

1

Retail Sales

Plan stock, purchasing and delivery around weekly, seasonal and holiday patterns.

2

Website Traffic

Forecast visits and usage to plan how much system capacity is needed.

3

Business Numbers

Forecast revenue, new customers and user growth for planning.

Retail demand changes throughout the year.

Sales are affected by weekly shopping habits, seasons, promotions, and holidays. These patterns directly affect inventory, purchasing, delivery planning, and revenue expectations.

A 2020 research study, Application of Facebook’s Prophet Algorithm for Successful Sales Forecasting Based on Real-world Data, used Prophet on actual sales data from a large retail company in Bosnia and Herzegovina.

The study showed how Prophet, together with testing on past data, can be used to forecast retail sales and check how reliable the forecast is for different products.

The goal is not only to predict sales.

It is to help the business make better decisions about stock availability, purchasing schedules, delivery planning, and cash planning.

A forecast is useful when it connects directly to a business decision.

Websites and online platforms produce operational data all the time.

Website visits, app usage, and system load often follow patterns based on weekdays, weekends, seasons, and business events.

Meta’s original work on Prophet aimed to make business forecasting easier to use at a larger scale, including forecasting the numbers used to plan how much system capacity is needed.

One example is using past website traffic to forecast future demand.

The aim is to understand how traffic is likely to change and whether the current infrastructure can support it.

For leaders, the benefit is linking the forecast to infrastructure spending decisions.

Instead of relying only on past peak usage or reacting once demand rises, teams can use forecast demand as one more input when planning capacity.

This is helpful when cost, system performance, and customer experience all need to be balanced.

Organizations regularly track numbers such as revenue, new customers, user activity, and operational volumes.

Prophet was designed to support business forecasting at larger scale, including numbers such as revenue and user growth.

These numbers often show repeating patterns, an overall growth trend, and changes caused by business events.

This is useful when forecasts need to be reviewed and understood by finance, operations, and leadership, and not only by the data team.

The forecast can support revenue planning, growth estimates, resource allocation, and operational reviews.

One thing to keep in mind: a forecast of a business number does not explain why that number is changing.

Understanding the effect of pricing, marketing campaigns, customer behavior, or other business factors may need additional analysis.

Prophet builds a forecast from parts of the past data: the trend, the seasonal patterns, and the effect of holidays and events. These parts reflect patterns that are common in business data, and they do not require a lot of manual setup.

What Prophet Looks At

The parts of past data that shape a Prophet forecast

1

Trend

The general direction of the numbers over time.

2

Weekly patterns

Repeating changes across the days of the week.

3

Yearly patterns

Repeating changes across the seasons of the year.

4

Holidays and events

Known dates, such as holidays or product launches, that affect the numbers.

From Past Data to a Business Decision

1

Past Business Data

Sales, traffic, revenue or other numbers.

2

Prophet

Looks at trend, seasonal patterns, holidays and events.

3

Forecast

Shown as trend and seasonal parts.

4

Test on Past Data

Check how close the forecast was.

5

Business Decision

Stock, capacity, planning.

For organizations, this offers several advantages.

Why Use Prophet

1

Easy to Understand

The forecast can be split into trend and seasonal parts that are easy to discuss.

2

Repeating Patterns

Weekly and yearly patterns can be included together.

3

Practical to Use

Available in Python and R, so analytics teams can use their current tools.

4

Imperfect Data

Works with missing data points and unusual values. Data quality still needs checking.

5

Many Numbers

Can be used for many business numbers or products. Check each forecast before using it.

Prophet should not be chosen only because it is well known.

In some situations, extra care is needed.

Prophet should be used only when it suits the business need.

The amount of past data, how far ahead you need to forecast, seasonal patterns, and the cost of a wrong forecast should all be considered.

Is Prophet a Good Fit?

Define the business need What decision will the forecast support?

↓

Review the situation Amount of past data, how far ahead to forecast, seasonal patterns, and the cost of a wrong forecast.

↓

Prophet is a practical fit Repeating seasonal patterns, changing trends, known holidays or events.

Use Prophet with extra care Short history, products that sell rarely, or results driven mainly by outside factors.

↓

Test the forecast on past data Check how close it was to what actually happened.

↓

Connect the forecast to a business decision

A forecast helps most when it connects past business data to future decisions, and is not treated as just another tool.

A good forecasting effort should answer four questions:

From Forecast to Business Action

1

What are we forecasting?

Revenue, demand, traffic, system usage, or another number the business can measure.

2

Why are we forecasting it?

Inventory planning, capacity planning, financial planning, or operational readiness.

3

How reliable is the forecast?

Test the forecast on past data and measure how far off it was.

4

What action will it lead to?

Link the forecast to decisions, workflows, and a clear owner in the business.

Without these steps, a forecast may be only a number on a dashboard that does not change how the organization works.

Prophet is a practical way to forecast business numbers that show repeating seasonal patterns, changing trends, and known calendar effects.

It is most useful for retail demand forecasting, planning website and system capacity, and tracking business numbers.

But Prophet should not be forced onto every forecasting problem.

The aim should be to find where a forecast can improve a decision, check how reliable it is, and build it into the business process.

The question is not only whether Prophet can forecast the future. It is whether the forecast helps the business make a better decision today.

That is how a forecast becomes useful to the business.